Technologies for extracting and analysing the credibility of health-related online content

Abstract

The evolution of the Web has led to an improvement in information accessibility. This change has allowed access to more varied content at greater speed, but we must also be aware of the dangers involved. The results offered may be unreliable, inadequate, or of poor quality, leading to misinformation. This can have a greater or lesser impact depending on the domain, but is particularly sensitive when it comes to health-related content. In this thesis, we focus in the development of methods to automatically assess credibility. We also studied the reliability of the new Large Language Models (LLMs) to answer health questions. Finally, we also present a set of tools that might help in the massive analysis of web textual content.